I type "US invade IRAN". My Claude runs:

- pulls my live positions, filters Iran cluster → know what I'm holding before reading anything else
- fetches market state (price / vol / liq / endDate / days left)
- 7d hourly history → spikes, drift, anomalies
- resolver text verbatim → flags ambiguity (ground troops only? airstrikes? UAE proxy seizing islands?)
- cluster term structure → every related Iran tenor side-by-side
- parallel OSINT: X via specific handles (not generic "Iran" search), TG aggregator channels 12h, operational data — Hormuz vessel counts, Brent curve, P&I insurance status, force majeures
- whale-intel: Tier S wallets on this cluster, distribute/accumulate events last 14d
- cluster delta scan: one tenor moved 5pp, adjacent didn't → mispricing flag
- frame check back to me: resolution bet or reprice scalp?
- calibration band: conservative / central / liberal, with assumptions that shift each
- where the agent's view and mine diverge — surfaced explicitly
Structure my agent runs when I type that. Lets me check:
1) resolver edge — am I right on the outcome math?
2) headline game — does this news move it or is it noise?
3) cluster arbitrage — one market moved, related didn't?
Copy it into your Claude and thank me later.